首页 时间轴 标签 归档 资源 关于

目录

嵌入整张图

 收录于 机器学习 Python
   约 134 字   预计阅读 1 分钟 

如何嵌入

全图进行编码嵌入

https://cdn.spiritlhl.net/https://github.com/spiritysdx/images/blob/main/20230921/13.png?raw=true

https://cdn.spiritlhl.net/https://github.com/spiritysdx/images/blob/main/20230921/14.png?raw=true

引入虚拟节点

https://cdn.spiritlhl.net/https://github.com/spiritysdx/images/blob/main/20230921/15.png?raw=true

匿名随机游走

https://cdn.spiritlhl.net/https://github.com/spiritysdx/images/blob/main/20230921/16.png?raw=true

https://arxiv.org/pdf/1805.11921.pdf

https://cdn.spiritlhl.net/https://github.com/spiritysdx/images/blob/main/20230921/17.png?raw=true

只要是新节点就标记新的一号，认号不认原来的节点标识。

https://cdn.spiritlhl.net/https://github.com/spiritysdx/images/blob/main/20230921/18.png?raw=true

https://cdn.spiritlhl.net/https://github.com/spiritysdx/images/blob/main/20230921/19.png?raw=true

使用不同匿名随机游走序列的个数构成的向量作为全图随机游走的向量。

https://cdn.spiritlhl.net/https://github.com/spiritysdx/images/blob/main/20230921/20.png?raw=true

给每种匿名随机游走的序列单独学习一个嵌入编码，全图也单独学习一个嵌入编码

https://cdn.spiritlhl.net/https://github.com/spiritysdx/images/blob/main/20230921/21.png?raw=true

https://cdn.spiritlhl.net/https://github.com/spiritysdx/images/blob/main/20230921/22.png?raw=true

总结

https://cdn.spiritlhl.net/https://github.com/spiritysdx/images/blob/main/20230921/23.png?raw=true

实际还有一种方法：分层聚类

https://cdn.spiritlhl.net/https://github.com/spiritysdx/images/blob/main/20230921/24.png?raw=true

https://cdn.spiritlhl.net/https://github.com/spiritysdx/images/blob/main/20230921/25.png?raw=true

https://cdn.spiritlhl.net/https://github.com/spiritysdx/images/blob/main/20230921/26.png?raw=true

更新于 2023-09-21
 机器学习, 深度学习, 图神经网络, DeepWalk, Node2Vec, CS244W
返回 | 主页